Ablation of the Retinoblastoma gene family deregulates G₁ control causing immortalization and increased cell turnover under growth-restricting conditions

Abstract

The retinoblastoma suppressor pRB belongs to the family of so-called pocket proteins, which also includes p107 and p130. These proteins may functionally overlap in cell cycle control and tumor suppression. We have generated an isogenic set of embryonic stem (ES) cell lines carrying single or compound loss-of-function mutations in the Rb gene family, including a cell line completely devoid of all three pocket proteins. None of the knockout combinations affected the growth characteristics of ES cells; however, concomitant ablation of all three pocket proteins strongly impaired their differentiation capacity. For the generated genotypes, primary mouse embryonic fibroblasts (MEFs) also were obtained. While inactivation of Rb alone did not alleviate the senescence response of MEFs, pRB/p107-deficient MEFs, after having adapted to in vitro culturing, continued to proliferate at modest rate. Additional ablation of p130 rendered MEFs completely insensitive to senescence-inducing signals and strongly increased their proliferation rate. Although triple-knockout MEFs retained anchorage dependence, they lacked proper G₁ control and showed increased cell turnover under growth-inhibiting conditions.

Loss of function of the retinoblastoma suppressor gene, RB, is a common event in the development of many tumor types in human, including hereditary retinoblastoma and sporadic lung, breast, and bladder carcinomas (Harbour et al. 1988; Lee et al. 1988; Horowitz et al. 1990). Numerous biochemical studies have placed pRB at the heart of the cellular machinery that controls passage from G₁ into S phase of the cell cycle (Weinberg 1995). pRB can exist in hyper- and hypophosphorylated forms, the latter binding to and inhibiting a class of transcription factors, collectively known as E2 factors (E2F), whose activity is required for the transcription of genes that are essential for DNA synthesis (Bernards 1997; Dyson 1998; Nevins 1998). Phosphorylation of pRB by cyclin-dependent kinases (CDK) leads to dissociation of the pRB/E2F complex, releasing the transcriptional activity of E2Fs. CDK activity is positively regulated by heterodimerization with cyclins but also negatively regulated by cyclin-dependent-kinase inhibitors (CKIs). pRB function can thus be described as a cell cycle switch: cyclin D1-stimulated kinase activity of CDK4 turns pRB -OFF- through phosphorylation; this releases E2F activity and promotes G₁-S transition; the pRB -ON- state is favored by inhibition of CDK4 activity by the CKI p16^INK4A, which promotes cell cycle arrest. The importance of this G₁-S control pathway is underscored by the finding that in the majority of human cancers, genetic alterations have been found that favor the pRB -OFF- state. These include genetic inactivation of RB in retinoblastoma and many other cancers; cyclin D1 overexpression in carcinomas of the breast, oesophagus, and head and neck; CDK4 amplification or mutational activation in melanomas; and abrogation of p16^INK4A activity in melanomas, pancreatic, and bladder carcinomas (Hall and Peters 1996; Sherr 1996).

Although the G₁-S control pathway described above provided a framework for understanding the tumor-suppressor function of pRB, cell cycle control in mammalian cells is a far more complex circuitry of different, at least partially overlapping and interacting pathways that are regulated by positive and negative feedback mechanisms and can be modulated at numerous levels. Thus, pRB shares its E2F-regulating activity with two homologs, p107 and p130 (for review, see Mulligan and Jacks 1998; Nevins 1998; Lipinski and Jacks 1999). The three proteins share extensive structural homology, most notably located in two regions, A and B, which together form the so-called pocket domain responsible for binding to E2Fs. These domains also form the binding site for many viral oncoproteins, including adenovirus E1A, simian virus 40 large-T antigen, and human papillomavirus E7, all of which abrogate interactions with E2F (DeCaprio et al. 1988; Whyte et al. 1988; Dyson et al. 1989). Also outside the A/B domain, p107 and p130 share extensive homology, whereas corresponding pRB regions and even the pRB spacer region that separates the A and B subdomains appear to be unique. The closer resemblance between p107 and p130 is underscored by a number of observations suggesting that the two proteins can functionally substitute for each other. p107 and p130 almost exclusively interact with E2F4,5, although they do so at consecutive stages of the cell cycle: Complexes of p130 and E2F4,5 predominate in quiescent or differentiated cells; p107/E2F4,5 complexes are only detected in cycling cells. In contrast, pRb has strongest affinity for E2F1,2,3, although it can also bind E2F4 (for review, see Dyson 1998). Rather than merely sequestering E2Fs, pocket-protein/E2F complexes appear to be actively involved in transcriptional repression (Zhang et al. 1999). While no differences in expression of E2F target genes were found on ablation of p107 or p130 alone, combined loss in p107^−/−p130^−/− primary mouse embryonic fibroblasts (MEFs) resulted in derepression of a number of genes, including B-myb, Cdc2, E2f1, Ts, Rrm2, and Cyclin A2. Expression of these genes was normal in Rb^−/− MEFs, but in these cells, Cyclin E and p107 were slightly up-regulated (Herrera et al. 1996; Hurford et al. 1997). Thus, p107 and p130 appeared to be redundant in regulating a specific set of E2F-responsive genes that is not subject to pRb/E2F-mediated repression. However, overexpression of each of the pocket proteins in vitro can cause cell cycle arrest, an activity that can be relieved through phosphorylation by cyclinD/CDK4,6. Full inactivation of pRB also requires cyclinE/CDK2, which adds an additional level of regulation to the pRB pathway. Thus, the three pocket proteins may to some extent be redundant in blocking progression of cells through the cell cycle. In resting T lymphocytes, the main pocket protein/E2F complex is p130/E2F4. However, in the absence of p130, E2F4 was mainly found complexed to p107 or, when this protein was also absent, to pRb (Mulligan et al. 1998). Apparently, in this system pocket proteins can compensate for each other in regulating E2F4 activity, providing a rationale for their shared activities in blocking cell cycle progression.

Specific and redundant functions of pocket proteins also became manifest in mice carrying single or compound disruptions in the Rb gene family (Mulligan and Jacks 1998). Complete inactivation of Rb appeared not to be compatible with normal development (Clarke et al. 1992; Jacks et al. 1992; Lee et al. 1992). Rb^−/− embryos died at midgestation, presenting with extensive apoptosis in the central nervous system and eye lens and defects in erythropoiesis and myogenesis. These defects could at least partially be attributed to increased free E2F1 activity, inducing p53-dependent apoptosis (Morgenbesser et al. 1994; Tsai et al. 1998). In contrast, genetic ablation of either p107 or p130 did not show an overt phenotype (Cobrinik et al. 1996; Lee et al. 1996). However, combination of p107 and p130 loss-of-function mutations caused excessive chondrocyte proliferation leading to malformation of the long bones and ribs. Respiratory problems associated with this defect led to rapid neonatal death. Remarkably, a dramatic phenotype was observed when loss-of-function mutations in either p107 or p130 were crossed into the BALB/c background (LeCouter et al. 1998a,b): p107 deficiency caused impaired growth of newborns, while embryonic fibroblasts displayed a twofold reduction in doubling time; p130-deficient embryos died at midgestation and showed marked apoptosis in the central nervous system. BALB/c mice were found to carry a point mutation in the p16^Ink4a gene (Zhang et al. 1998). These observations suggest that p107 and p130 can functionally substitute for each other, but only in the presence of a pathway that requires wild-type p16 and most likely involves pRB. Conversely, p107 appeared to attenuate the consequences of pRB deficiency in development as p107/pRB-deficient embryos showed a reduced life span with respect to embryos lacking only pRB (Lee et al. 1996). p107 activity also appeared to modulate the tumor suppressor function of pRB. In contrast to humans, where hemizygosity for the retinoblastoma suppressor gene RB highly predisposes to retinoblastoma, Rb^+/− mice developed tumors originating from the intermediate lobe of the pituitary gland while the retina remained unaffected (Hu et al. 1994; Robanus- Maandag et al. 1994). No predisposition to tumorigenesis was found to be associated with p107 or p130 deficiency (Cobrinik et al. 1996; Lee et al. 1996). However, concomitant inactivation of Rb and p107 in chimeric mice gave rise to retinoblastoma development, indicating that p107 can act as a tumor suppressor in the context of pRb deficiency (Robanus-Maandag et al. 1998).

Primary mouse embryonic fibroblasts represent a rather well-defined cell type that has been widely used to identify the consequences of gene ablation for cell cycle control. Although pRB deficiency or combined p107/p130 deficiency caused a slight acceleration of S-phase entry on restimulation of serum-starved cells, the overall growth characteristics of single- or double-knockout MEFs did not deviate from those of wild-type MEFs (Cobrinik et al. 1996; Herrera et al. 1996). In particular, the single- and double-knockout genotypes tested thus far did not affect the growth arrest response of cells to a number of growth-inhibiting conditions such as serum deprivation, contact inhibition, and prolonged passaging. This again suggests functional redundacy within the pocket-protein family. To test this directly, we have now generated MEFs that are completely devoid of all three pocket proteins. Our results identify the pocket proteins as critical mediators of senescence and of the response of cells to a variety of growth-restricting conditions.

Results

Inactivation of Rb, p107, and p130 in ES cells

An isogenic set of wild-type (wt), Rb^−/−, Rb^−/−p107^−/−, and Rb^−/−p107^−/−p130^−/− embryonic stem (ES) cell lines was generated by consecutive rounds of homologous recombination. The generation of Rb^−/− and Rb^−/−p107^−/− ES cell lines has been described before (Te Riele et al. 1992; Robanus-Maandag et al. 1998). To inactivate p130, we generated a 129p130-pur targeting vector (Fig. 1A). Rb^−/−p107^−/−p130^−/− cells were obtained in two ways: First, p130 was inactivated in Rb^−/−p107^−/− ES cells and Rb^−/−p107^−/−p130^+/− cells were cultured at high puromycin concentrations to select for inactivation of the remaining p130 wild-type allele; second, both p130 alleles were inactivated in p107^−/− ES cells and then Rb was inactivated using the targeting vectors 129Rb-hyg and 129Rb-his. Figure 1B shows that p130 protein was not detectable in brain tissue obtained from a p130^−/− mouse carrying the same disruption in p130, indicating that the p130⁻ allele can be considered a null allele. Thus, Rb^−/− p107^−/−p130^−/− cells are completely devoid of pocket proteins and will hereafter be designated as triple-knockout (TKO) cells. Each genotype was derived in duplicate by independent rounds of gene targeting starting from wild-type cells. None of the generated genotypes in ES cells resulted in an obvious phenotype, indicating that the pocket proteins pRB, p107, and p130 are dispensable for normal ES cell growth.

Figure 1.

Generation of Rb^−/−p107^−/−p130^−/− ES cells and MEFs. (A) Restriction map of the wild-type p130 allele around codon 405 at the AvaI site and 129p130-pur DNA targeting construct. Probes 1 and 2 detect modifications at p130. (B) Western blot analysis of p130 in lysates prepared from p130^+/+ and p130^−/− brain tissue. (C) PCR analysis of Rb, p107 and p130 of DNA isolated from passage 1 wt, Rb^−/−, Rb^−/−p107^−/− and TKO MEFs. PCR products resulting from wild-type and knockout alleles of each gene are indicated by open and solid arrows, respectively. (D) Western blot analysis of pRB, p107, p130, and Cyclin E in lysates prepared from indicated MEFs. Cdk4 served as a loading control.

Limited differentiation capacity of Rb^−/−p107^−/−p130^−/− ES cells in teratocarcinomas

To study whether pocket proteins become essential during differentiation of ES cells, we injected wt, Rb^−/−, Rb^−/−p107^−/−, and Rb^−/−p107^−/−p130^−/− ES cells subcutaneously into nude mice. In this environment, ES cells form teratocarcinoma-like tumors in which a variety of differentiated cell types can be identified, including neuronal and muscle cells. Three weeks after injection the teratocarcinomas were removed and histologically examined. Tumors derived from wt, Rb^−/−, and Rb^−/−p107^−/− ES cells were highly similar with respect to their strongly heterogeneous appearance with extensive neuronal differentiation (Fig. 2A) and striated muscle (Fig. 2B). Extensive neuronal differentiation was confirmed by staining with the neuron-specific antibodies anti-glial fibrillary acidic protein (GFAP) and anti-neuron specific enolase (NSE; not shown). In contrast, TKO ES cells formed a rather homogeneous tumor mass, without any muscle cell differentiation (Fig. 2C). Rosette-like structures and positive staining with a p75^LNGFR antibody (Zorick and Lemke 1996) indicated that TKO tumors predominantly consisted of neuroblast-like cells (Fig. 2D; not shown). Furthermore, immunohistochemistry using antibodies directed against the proliferation marker Ki-67 showed that the fraction of proliferating cells in TKO tumors was 15-fold higher than in teratocarcinomas of the other genotypes. These data show that in this system, ablation of the Rb gene family abrogates neuronal and muscular differentiation.

Figure 2.

Limited differentiation in TKO teratocarcinomas. Histological sections of teratocarcinomas were stained with hematoxylin-eosin. (A,B) Advanced neuronal (N) and muscle (M) differentiation in teratocarcinomas generated with wt ES cells. Similar results were obtained with Rb^−/− and Rb^−/−p107^−/− ES cells. (C,D) Cell-dense TKO teratocarcinomas lack muscle differentiation and exist of primitive, relatively undifferentiated neuronal cells (PN) with rosette-like structures (R). Magnification (A,C) objective 10×, (B,D) objective 20×.

Pocket-protein-deficient MEFs obtained from chimeric embryos

Primary MEFs have been instrumental in studying the consequences of gene ablation for cell cycle control. We have derived an isogenic set of MEFs from chimeric embryos generated by blastocyst injections of mutant ES cells. In these experiments, we used ES cells that also contained a neomycin resistance gene under the control of the Rosa26 promoter (Friedrich and Soriano 1991; J.-H. Dannenberg and H. te Riele, in prep.). Brief culturing in the presence of G418 provided a uniform method to obtain MEF preparations that were exclusively ES-cell derived. This was verified by genotyping these cells by PCR for the presence of the wild-type and knockout alleles of Rb, p107, and p130 (Fig. 1C). The presence or absence of pocket proteins was further confirmed by Western blot analyses (Fig. 1D). We were able to generate MEF cultures with the following genotypes: wt, Rb^−/−, Rb^−/− p107^−/−, and somewhat surprisingly, Rb^−/−p107^−/−p130^−/−.

Pocket proteins are essential for G₁ control

Proliferating MEF cultures of early passage were labeled with the thymidine analogue BrdU and analyzed by FACS. Figure 3A shows that concomitant inactivation of Rb, p107, and p130 strongly reduced the percentage of cells in G₁. This decrease in G₁ fraction was accompanied by an increase of the S-phase population, with TKO MEF cultures having up to 65% S-phase cells. In Rb^−/− MEFs, only a minor decrease in G₁ population was observed, whereas disruption of both Rb and p107 gave an intermediate phenotype. Together with the strongly reduced population-doubling time of TKO cells (see below; Fig. 4), these observations suggest that successive ablation of pRb, p107, and p130 reduces the length of the G₁ phase of the cell cycle.

Figure 3.

Deregulated G₁-control in TKO MEFs. (A) Percentages of G₁, S, and G₂/M phases of the cell cycle in proliferating wild-type (wt) and pocket-protein-deficient MEFs. Results are indicated as average values of three independent experiments. Error bars indicate standard deviations of the mean. (B) FSC-H histogram showing cell-size analysis of MEFs deficient for Rb, p107, and p130 compared with wt counterparts. The forward scatter, FSC-H, is indicative for the cell diameter, and a shift to the left, relative to the wt MEFs, represents smaller cells.

Figure 4.

Growth characteristics of pocket-protein-deficient MEFs. (A) Growth curves of wt, Rb^−/−, Rb^−/−p107^−/−, and TKO MEFs at passages 3 (squares), 5 (circles), and 7 (triangles). The figure shows a representative of three independent experiments, each performed in triplicate. (B) Cell proliferation on a 3T9-based protocol of indicated MEFs for 20 passages. At 3.5-d intervals, the total numbers of cells of three independent cultures of each genotype were determined before redilution of the cells to 3 × 10⁵ per six-well plate (input). The experiment was performed twice. Error bars indicate standard deviations of the mean.

G₁ shortening has previously been correlated to a decrease in cell size in several other experimental settings, such as overexpression of cyclin E (Ohtsubo and Roberts 1993) and concomitant inactivation of Rb and p21 (Brugarolas et al. 1998). We therefore determined the size of wild-type and mutant MEFs. Proliferating cultures of early passage were fixed, stained with propidiumiodide (PI), and analyzed by FACS (Fig. 3B). Relative cell sizes closely correlated to the proportion of cells in G₁: TKO<Rb^−/−p107^−/−<Rb^−/− = wt. This again indicates that ablation of pocket proteins reduces the length of G₁.

It is of interest to note that in cyclin-E-overexpressing or Rb/p21-deficient cells, G₁ shortening has been correlated with increased cyclin E levels or elevated CDK2 kinase activity. However, whereas Rb^−/− and Rb^−/− p107^−/− MEFs showed increased levels of cyclin E, TKO MEFs expressed cyclin E at wild-type levels (Fig. 1D). Thus, shortening of G₁ is not obligatorily accompanied by elevated levels of cyclin E.

Rb^−/−p107^−/−p130^−/− MEFs are immortal

Wild-type MEFs have a limited growth capacity on prolonged passaging: Their growth rate gradually decreases until they finally completely arrest with an enlarged, flattened morphologoly and 2n DNA content, a state that is referred to as replicative senescence (Todaro and Green 1963; Campisi 1997). The involvement of the Rb gene family in senescence was assessed in two ways. First, we determined the growth rates of wt, Rb^−/−, Rb^−/− p107^−/−, and TKO MEFs at increasing passages. Figure 4A shows that the growth rates of wt and Rb^−/− MEFs gradually decreased on prolonged passaging and that these cells entered senescence around passage seven. Rb^−/− p107^−/− MEFs showed a fivefold reduction in growth rate; however, they had not completely ceased proliferating at passage seven (Fig. 4A) and clearly had not adopted a senescence-like morphology (not shown). TKO MEFs retained the strongest proliferative capacity at later passages, their growth rate being reduced less than twofold. Second, the long-term proliferative capacity of MEFs was assessed following a 3T9-based protocol (Fig. 4B). Wild-type, Rb^−/−, and Rb^−/−p107^−/− MEFs showed a decline in proliferation rate up to passage 11. While the Rb^−/− cells completely stopped growing in this experiment, wt cells ultimately escaped from senescence as a result of a genetic alteration (see also below). Rb^−/− p107^−/− MEFs, although initially subject to growth inhibition, did not enter senescence but established a constant proliferation rate of one to two population doublings every 3 d. Also, TKO cells initially showed some decline in growth rate. However, this decline was marginal, and the cells continued to proliferate at a rate that was about threefold higher than the double knockouts. These results demonstrate that inactivation of Rb alone does not alleviate the senescence response of MEFs; concomitant inactivation of Rb and p107, although not completely alleviating growth inhibition on in vitro culturing, confers a continuing proliferation capacity, albeit at a modest rate, whereas ablation of all Rb-gene family members renders cells virtually insensitive to growth inhibition on in vitro culturing and strongly increases their proliferation rate.

Immortal TKO MEFs retain an intact p19^ARF/p53 pathway

It has previously been hypothesized that replicative senescence results from the p53-dependent growth inhibitory effect of p19^ARF, which gradually accumulates on prolonged passaging (Quelle et al. 1995; Zindy et al. 1997). Thus, inactivation of Ink4a, p19^Arf, or p53 has been shown to result in immortalization of MEFs (Serrano et al. 1996, 1997; Kamijo et al. 1997). Furthermore, spontaneous immortalization of MEFs was accompanied by either mutation of p53 (Harvey and Levine 1991; Rittling and Denhardt 1992) or deletion of the Ink4a locus (Kamb et al. 1994; Nobori et al. 1994; Zindy et al. 1998), ablating both p16^INK4A and p19^ARF (Kamijo et al. 1997). We therefore tested the functionality of this pathway in immortal TKO MEFs. Figure 5A shows that p16^Ink4a and p19^Arf were properly expressed in all genotypes including TKO and tended to increase on prolonged passaging. Also, p53 function was induced as evidenced by increased protein levels and accumulation of p21^CIP (Fig. 5B). Furthermore, when TKO MEFs were γ irradiated, they showed a strong induction of p21^Cip, as in wt and Rb^−/−p107^−/− MEFs, again indicating that the cells contained functional p53 (Fig. 5C). These observations indicate that complete loss of function of the Rb gene family renders MEFs resistant toward p19^ARF/p53-induced senescence and predicts that TKO MEFs at passage numbers where wild-type MEFs spontaneously escape from senescence can grow without mutations in p53 or deletions of the Ink4a locus. Indeed, we found that in six independent clones that were grown on a 3T9-based protocol for >20 passages, p19^ARF was still expressed, in contrast to a wt 3T9 clone that had lost p19^ARF expression, probably because of a deletion of the Ink4a locus (Fig. 5D). Finally, all six TKO 3T9 clones showed increased p21^Cip expression on γ irradiation, as in wt and TKO MEFs of early passage, indicating that p53 also remained functional in these clones (Fig. 5E).

Figure 5.

Intact p19^ARF/p53 pathway in immortal TKO MEFs. (A) Indicated MEFs were passaged according to a 3T9 protocol, and cell pellets were collected at passages 3, 5, and 7. Cell lysates were analyzed for p16^INK4A, p19^ARF and tubulin levels by immunoblotting. Ink4a^−/− MEFs served as a negative control. (B) Expression of p53 and p21^CIP in wt and TKO MEFs at passages 3 and 6. (C) Expression analysis of p21^CIP in control and γ-irradiated early passage wt, Rb^−/−p107^−/− and TKO MEFs showing induction of p21^Cip in all genotypes 24 h after irradiation. (D) Expression analysis of p19^ARF in six independent TKO 3T9 clones (1–6). All clones expressed p19^Arf, whereas a wild-type 3T9 clone (3T9wt, 1) had lost expression of this protein. p19^Arf−/− MEFs served as a negative control. (E) Induction of p21^Cip in γ-irradiated wt (1) and TKO (1–6) 3T9 MEF cell lines compared to γ-irradiated early passage wt and TKO MEFs.

p53 can induce both a pRB-dependent G₁ and a pRB-independent G₂/M arrest on γ irradiation (Agarwal et al. 1995; Brugarolas et al. 1995; Harrington et al. 1998). Figure 6A shows that a γ-irradiation-induced G₁ arrest was indeed abrogated in Rb^−/−p107^−/− and TKO MEFs, while the G₂/M checkpoint was still intact. This again indicates that in Rb^−/−p107^−/− and TKO cells, p53 remained functional and that pocket proteins are dispensable for radiation-induced G₂/M arrest.

Figure 6.

Defective G₁ checkpoint in TKO MEFs on irradiation and ectopic p19^Arf expression. (A) γ-irradiation induced a G₂/M arrest in Rb^−/−p107^−/− and TKO MEFs. Percentages of G₁, S, and G₂/M phases of the cell cycle in nonirradiated and irradiated (5.5 Gy) MEFs were measured by FITC-PI FACS analysis and plotted as FL3-A (X-axis)–FL1-H (Y-axis) dot plots. (B) Ectopic expression of p19^Arf induces a pocket-protein-dependent G₁ arrest. FACS analysis of GFP-positive Rb^−/−, Rb^−/−p107^−/− and TKO MEFs expressing either vector plasmid, a dominant-negative variant of CDK2, or p19^Arf (1.0 or 5.0 μg) together with a histone H2B-GFP expression construct in a 10 : 1 ratio. Representative results of three independent experiments are shown.

Finally, we asked whether MEFs deficient for the Rb-gene family are still sensitive to G₁ arrest following ectopic expression of p19^ARF. Cells were electroporated and analyzed by FACS. First, we found that introduction of a CDK2 dominant-negative protein increased the G₁ population to the same extent in Rb^−/−, Rb^−/−p107^−/−, and TKO MEFs (Fig. 6B). Ectopic expression of p19^ARF induced a similar G₁ arrest in cells lacking pRB or both pRB and p107. In contrast, in TKO MEFs, G₁ arrest was only marginally induced by p19^ARF compared to the effect of CDK2^dn (Fig. 6B). Together with the observation that ablation of the Rb gene family causes immortalization despite normal levels of INK4A proteins and functional p53, this result indicates that pocket proteins are essential mediators of the senescence response to p19^ARF.

Inactivation of pocket proteins leads to increased cell turnover under growth-inhibiting conditions

TKO cells are largely refractory to growth inhibition by p19^ARF. We therefore examined the behavior of these cells under a number of other growth-inhibiting conditions.

Growth factor depletion

MEFs were plated at low density, supplied with 0.1% serum, and analyzed by BrdU-FACS. Wt and Rb^−/− cells strongly arrested as evidenced by the accumulation of cells in G₁ and a block in DNA synthesis (Fig. 7A). The response of Rb^−/−p107^−/− MEFs was less pronounced, although the level of DNA synthesis was still significantly decreased compared to 10% serum (see also Fig. 3A). In contrast, TKO MEFs continued to incorporate BrdU almost at the same rate as under high serum conditions. Suprisingly, under these conditions, no increase in cell number was observed on prolonged culturing (Fig. 7B). The appearance of floating cells in the medium at 24 h after the application of low serum suggested that cells became apoptotic. Indeed, PI-FACS analysis revealed a significant sub- G₁ population in TKO MEFs at 24 h after low serum addition, increasing up to 10% of the population after 4 d (Fig. 7C). Again, Rb^−/−p107^−/− showed an intermediate phenotype. We did not observe an increased population of multinucleated MEFs, indicating that continuous BrdU incorporation did not result from endoreduplication (data not shown).

Figure 7.

Increased cell turnover in TKO MEFs under growth-inhibiting conditions. (A) Distribution of G₁, S, and G₂/M phases of indicated MEFs at 0.1% FCS. (B) Growth curves of indicated MEFs at 0.1% FCS. (C) Percentage of apoptotic cells in indicated MEF cultures at 0.1% FCS, 24 h (black), and 96 h (grey) after growth-factor depletion. Percentages were obtained by determing the sub-G₁ population in a FL3A-log histogram. (D) Distribution of G₁, S, and G₂/M phases of indicated contact-inhibited MEFs. (E) Percentage apoptotic cells in contact-inhibited MEF cultures. Average values of three independent experiments are indicated in percentages. Error bars indicate standard deviations from the mean (A,C,D,E).

Contact inhibition

Another extracellular signal that provokes a G₀/G₁ arrest in normal fibroblasts is cell–cell contact at confluency (Nilausen and Green 1965). This phenomenon, known as contact inhibition, was shown to be accompanied by increased p27^KIP1 and decreased cyclin D1 levels (Polyak et al. 1994; St. Croix et al. 1998). Furthermore, hypophosphorylated isoforms of pRb accumulated, suggesting that pRB is essential for G₁ arrest. To study the role of pocket proteins in contact inhibition, MEFs were grown to confluency and cultured for an additional 4 d. DNA synthesis was measured by BrdU-FACS. Again, deletion of all pocket proteins revealed a severe defect in establishing a G₀/G₁ arrest (Fig. 7D). Instead, TKO MEFs continued to incorporate BrdU and to generate apoptotic cells (Fig. 7E).

Nonadherent conditions

Anchorage independance is a characteristic of most oncogenically transformed cell types and was attributed to constitutive cyclin E- (Fang et al. 1996) or A- (Guadagno et al. 1993) dependent CDK2 activity. In nonmalignant cells, growth arrest under nonadherent conditions was associated with loss of cyclin E-CDK2 activity (Fang et al. 1996) and underphosphorylation of pRB and p107 (Schulze et al. 1996). However, we found that TKO MEFs, although giving very tiny colonies containing up to 20 cells, were not capable of sustained proliferation in soft agar (Table 1). This indicates that ablation of pocket-protein activity is not sufficient for growth under nonadherent conditions and suggests that loss of CDK2 activity can confer a pocket-protein-independent growth arrest. This view is consistent with our observation that TKO cells can still be blocked by ablation of CDK2 activity (Fig. 6B). The anchorage dependence of TKO MEFs is consistent with their inability to grow under the skin of nude mice (not shown), demonstrating that these cells are not transformed.

Table 1.

Immortal TKO MEFs do not grow anchorage independently

Genotype MEFs	Exp I	Exp II	Exp III
wt	0	0	0
Rb−/−	0	0	0
RB−/−p107−/−	0	0	0
Rb−/−p107−/−p130−/−	0a	0	0
p53−/− + RASV12	229	238	273

Number of foci of indicated early passage MEFs appearing after 2 wk in soft agar. 

^aA small percentage of the MEFs developed into tiny colonies (up to 20 cells) but did not sustain growth. 

In conclusion, complete ablation of the Rb gene family does not allow anchorage-independent growth; however, it severely compromises cell cycle arrest in response to growth-factor depletion and contact inhibition. Under these growth-inhibiting conditions, pocket-protein-deficient cells undergo increased cell turnover.

Discussion

We have generated an ES cell line carrying inactivating disruptions in all three retinoblastoma gene family members and found that this strongly limited their differentiating capacity. In MEFs, complete ablation of the pocket-protein family severely compromised G₁ control. This was most strikingly manifested by the finding that triple-knockout MEFs were immortal and sustained continuous proliferation and apoptosis under growth-restricting conditions.

Differentiation

ES cells have a very high proliferation rate (doubling time of 7–8 h), and in a proliferating population, ∼70% of the cells were found in S phase (Savatier et al. 1994). Savatier et al. (1996) made the surprising observation that undifferentiated ES cells do not contain detectable cyclin D1/CDK4 activity and are insensitive to growth inhibition by p16^INK4A. Furthermore, ES cells do not express p130 (LeCouter et al. 1996), whereas pRB and p107 could be readily detected (Robanus-Maandag et al. 1998). Apparently, in ES cells the antiproliferative effects of pRB and p107 are neutralized, possibly by the presence of an E1A-like activity (E1A-LA), as was postulated before (Murray et al. 1991). The dramatic increase in cyclin D1/CDK4 activity on induction of differentiation likely reflects the pRB/p107/p130 control pathways becoming active. Our observations are fully consistent with this view: Ablation of the pRB family did not affect the growth potential of ES cells but severely compromised their differentiating potential. Thus, pocket proteins are required for differentiation of ES cells. Moreover, as both pRB/p107-deficient ES cells (this work) and p130-deficient ES cells (not shown) showed normal muscular and neuronal differentiation in teratocarcinomas, p130 can compensate for the absence of pRB and p107.

Despite their limited differentiation capacity when placed under the skin of nude mice, TKO ES cells injected into blastocysts did not show unrestricted proliferaton but, instead, were capable of contributing to embryonic development at least until day 15 of gestation, allowing the derivation of TKO MEFs. This observation may suggest that differentiation into certain cell types, for example, fibroblasts, does not require pocket-protein function. More intriguing is the possibility that differentiation of TKO cells in chimeric embryos is directed and controlled by surrounding wild-type cells. In either case, during development, the cells have apparently acquired characteristics that, unlike their embryonic stem cell progenitors, prevent them from growing anchorage independently or under the skin of nude mice.

Immortalization

In both human and mouse primary fibroblasts, prolonged culturing generates a growth-inhibiting signal that ultimately leads to a state of replicative senescence reflected by an enlarged, flattened morphology and the absence of DNA synthesis (Hayflick and Moorhead 1961; Todaro and Green 1963; Campisi 1997). This type of growth arrest is accompanied by gradually increasing levels of the CDK2/4 inhibitors, p21^CIP and p16^INK4A, the cell cycle inhibitor p19^ARF, and p53 (Lloyd et al. 1997; Palmero et al. 1997; Zindy et al. 1997, 1998). p16^INK4A and p19^ARF are encoded by one genetic locus, Ink4a, whereby p19^ARF is expressed from an alternative reading frame (Quelle et al. 1995). Whereas p16^INK4A was shown to act upstream of pRB to promote cell cycle arrest (Serrano et al. 1993), p19^ARF can physically interact with p53 and/or MDM2, thereby antagonizing the function of MDM2 and stabilizing p53 (Kamijo et al. 1998; Pomerantz et al. 1998; Zhang et al. 1998). Spontaneous immortalization of MEFs is usually accompanied by either deletion of the Ink4a locus (Kamb et al. 1994; Nobori et al. 1994; Kamijo et al. 1997; Zindy et al. 1998) or loss of p53 function (Harvey and Levine 1991; Rittling and Denhardt 1992). Furthermore, Ink4a^−/− and p53^−/− MEFs are immortal. These observations suggest that induction of INK4A proteins and p53 is causally related to induction of senescence. However, mice and MEFs lacking only p19^ARF showed a phenotype indistinguishable from that resulting from disruption of the Ink4a locus abrogating both p16^INK4A and p19^ARF. Although the consequences of ablation of p16^INK4A alone are unknown, this suggests that the p19^ARF/p53 rather than the p16^INK4A/pRB pathway is responsible for inducing senescence.

We now show that MEFs deficient for pRB, p107, and p130 are immortal, cannot be arrested by ectopic p19^ARF expression, and do not sustain deletions of the Ink4a locus or mutations in p53 on prolonged passaging. Also, pRB/p107-deficient MEFs can be considered immortal: Although they respond to supraphysiological levels of p19^ARF by a G₁ arrest, after having adapted to in vitro culturing, they sustained a moderate but constant proliferation rate on prolonged passaging. These observations indicate that both p19^ARF-induced senescence and cell cycle arrest require pRB family members. It therefore remains plausible that both products of the Ink4a locus are implicated in senescence induction, p16^INK4A not being sufficient but facilitating the activation of pocket proteins by p19^ARF (Carnero et al. 2000). How p19^ARF signals to the pRB family remains unclear. The immortal phenotype of p53^−/− MEFs suggests this pathway to be p53 dependent. However, p19^ARF-p53-induced senescence is not implemented via p21^CIP-induced inhibition of cyclin E/CDK2, as p21^−/− and p21^−/−p27^−/− MEFs still undergo senescence and/or are responsive to overexpression of p19^ARF (Pantoja and Serrano 1999; Groth et al. 2000). Furthermore, our data show that TKO MEFs are not blocked in G₁ by p19^ARF overexpression, although they can still be blocked by inhibition of CDK2 activity on expression of a dominant-negatively acting CDK2 mutant. A picture thus emerges wherein p19^ARF induces senescence via the pRB family but independent of p21^CIP-mediated inhibition of cyclin E/CDK2 activity. Whether the link between p19^ARF and the Rb gene family is p53 dependent remains obscure. In contrast to one report (Kamijo et al. 1997), another showed that p19^ARF could induce a cell cycle arrest in p53^−/− MEFs, which can be relieved by overexpression of E2F-1 or by blocking p16^INK4A function (Carnero et al. 2000). These data suggest that p19^ARF can target the pRB pathway independent of p53. The only known target of p19^ARF is MDM2, which can bind p53 but also pRB (Xiao et al. 1995) and may, thus, be a candidate to mediate the p19^ARF/pRB connection.

The role of pocket proteins in oncogenic transformation

Expression of so-called transforming oncogenes such as oncogenic RAS (Palmero et al. 1998) and v-Abl (Radfar et al. 1998) strongly increases the level of p19^ARF. This activates a p53-dependent checkpoint that protects cells from abnormal mitogenic signaling by inducing replicative arrest. MEFs lacking p19^ARF or p53 function are no longer susceptible to this type of protection; they escape from replicative senescence and become oncogenically transformed (Harvey and Levine 1991; Harvey et al. 1993; Kamijo et al. 1997). Overexpression of another class of oncogenes such as MYC, adenovirus E1A, and E2F-1 in MEFs has been shown to enhance proliferation but also apoptosis (Lowe and Ruley 1993; Hermeking and Eick 1994; Wagner et al. 1994; Qin et al. 1994; Querido et al. 1997). Apoptosis is strongly enhanced by depriving cells of extracellular survival factors (Evan et al. 1992; Lowe and Ruley 1993) and is dependent on a functional p19^ARF-p53 pathway that does not involve p21^CIP function (Lowe et al. 1994; Wagner et al. 1994; Bates et al. 1998; De Stanchina et al. 1998; Zindy et al. 1998). The behavior of MYC- or E1A-expressing cells is highly reminiscent of that of triple-knockout cells: Pocket-protein ablation increased the proliferative capacity of cells but also their sensitivity to apoptosis: in particular, under conditions of low serum or high density. The picture that now emerges is that p19^ARF acts as a sensor of abnormal or conflicting mitogenic signaling, and activates a p53-dependent response that can either cause cell cycle arrest or sensitize cells to apotosis. The behavior of triple-knockout cells indicates that this decision depends on pocket-protein functions. In their presence, cells arrest; in their absence, for example, by genetic ablation, sequestration by E1A, or inhibition following overexpression of MYC (Berns et al. 2000; Lasorella et al. 2000), cells become immortal but also highly sensitive to apoptosis.

Indeed, we observed that culturing of triple-knockout MEFs under growth-restricting conditions readily induced apoptosis. At the same time, however, continuing cell proliferation was observed. Thus, under certain growth-restricting conditions, loss of pocket-protein function allows continuous cell turnover without net increase in cell number. We envisage that loss of pocket-protein function in vivo may also lead to increased cell turnover, which would strongly facilitate the acquisition of additional oncogenic events. Several observations lend support to this scenario. For example, conditional inactivation of Rb in the mouse pituitary gland resulted in increased cell turnover. Tumors appeared to arise from cells in which apoptosis was blocked, presumably through an additional mutation (Vooijs et al. 1997). Also, inactivation of the pRB family members in the mouse brain epithelium by the large T variant T121 induced aberrant proliferation and p53-dependent apoptosis. Inactivation of p53 in this setting resulted in a dramatic reduction in apoptosis (Yin et al. 1997). We speculate that redundancy of pocket proteins in cell cycle control is cell-type specific. For example, in the mouse pituitary gland, p107 and p130 may not be expressed or functional in compensating for loss of pRB, whereas in the mouse retina both pRB and p107 are active and, therefore, loss of both proteins is required to sustain prolonged cell turnover ultimately leading to retinoblastoma development (Robanus-Maandag et al. 1998).

In conclusion, we have identified three characteristics associated with loss of pocket-protein function that may provide a rationale for its frequent involvement in human cancer. These are reduced differentiation capacity, loss of the senescence response, and increased cell turnover under growth-restricting conditions. We expect that a careful study of the fate of ES cells carrying single or compound loss-of-function mutations in the Rb gene family in chimeric mice will provide further insight into the role of pocket protein ablation in tumorigenesis.

Materials and methods

Generation of ES cells deficient for pRb, p107, and p130

Knockout alleles of pRb, p107, and p130 were generated in 129Ola (E14/IB10) embryonic stem cells. The targeting vectors 129Rb-hyg, 129Rb-his, and 129p107-IRESβgeo have been described (te Riele et al. 1992; Robanus-Maandag et al. 1998). To inactivate p130, we generated the targeting construct 129p130-pur, in which a PGK-puromycin cassette was introduced into codon 405 residing in the A-domain of the pocket region (LeCouter et al. 1996). On electroporation and selection at 1.8 μg/mL of puromycin, homologous recombinants were obtained with a frequency of 80%. The remaining allele of p130 was inactivated by selection at high puromycin concentrations (14 μg/mL). p130^−/− mice were generated using chimeric mice which gave germ-line transmission.

Generation of MEFs deficient for pRb, p107, and p130

Chimeric embryos were generated by injection of mutant ES cells into a 3.5-d-old C57Bl/6 blastocyst. Embryos we isolated 12 d after implantation and put into culture. Briefly, embryonic tissue without organs was washed in PBS, minced and incubated in 100 μL trypsin/EDTA overnight on ice and then 30 min at 37°C with an additional 100 μL of trypsin/EDTA. Mutant MEFs were selected for 48 h with 800 μg/mL G418. These cells were designated passage 1. Nonchimeric MEF cultures, serving as a control, all died within 48 h of selection.

ES cell culture, generation and characterization of teratocarcinomas

ES cells were cultured as described (Robanus-Maandag et al. 1998). Nude mice were injected subcutaneously with 10⁶ ES cells. Tumors were removed 4 wk after injection, fixed in phosphate-buffered formalin, embedded in paraffin, sectioned at 5 μm, and stained with hematoxylin and eosin according to standard procedures. For immunohistochemical characterization, rehydrated sections were stained with antibodies against glial fibrillary acidic protein (GFAP, DAKO), low-affinity nerve growth factor receptor (p75^LNGFR, Chemicon International), neuron specific enolase (NSE, DAKO), and proliferation marker Ki-67 (MIB-5, Immunotech). The fraction of proliferating cells was determined by counting Ki-67-positive cells in six different areas in a teratocarcinoma section at a 160× magnification.

Culturing of MEFs, growth curves, 3T9 protocol

MEFs were cultured in GMEM (GIBCO) supplemented with 10% fetal calf serum, 1 mM nonessential amino acids, 1 mM sodium pyruvate, and 0.1 mM β-mercaptoethanol. For growth curves, MEFs were seeded at 2.5 × 10⁴ cells per 12-well plate, in triplicate. At various time points, cells were washed with PBS, fixed for 5 min in 4% formaldehyde, and stained with 0.1% crystal violet (Sigma) in distilled water for 30 min. After washing the cells two times with water, they were allowed to dry. Staining was retrieved from the cells by adding 1 mL 10% acetic acid per well. Optical density was measured of 100 μL retrieved staining at 590 nm. Values were normalized to the optical density at day 0.

A 3T9 protocol was performed by seeding 3 × 10⁵ cells per 6-well plate, in triplicate. Cultures were incubated at 37°C for 3.5 d, and the total amount of cells was determined using a cell counter (Casy 1) followed by replating 3 × 10⁵ cells. This protocol was repeated 20 times.

Electroporation of MEFs

MEFs were grown in 10-cm dishes to subconfluency. One to 2 × 10⁶ cells were resuspended in 100 μL electroporation buffer (2 mM Hepes at pH 7.2, 15 mM K₂HPO₄/KH₂PO₄, 250 mM mannitol, and 1 mM MgCl₂) at 37°C. The cell suspension was mixed with plasmid DNA (histone H2B-GFP expression plasmid [Kanda et al. 1998], together with either a dominant-negative variant of CDK2 [Van den Heuvel and Harlow 1993] or pcDNA-p19^Arf in a 1 : 10 ratio) and transferred to an electroporation cuvette (0.1 cm, Biorad). After 5 min of incubation at room temperature, the cells were electroporated with Gene Pulser II/Gene Pulser II RF module apparatus (Biorad). Five minutes later, electroporation cells were taken up in 8 mL 37°C medium and seeded into a 10-cm dish. To measure effects on G₁, electroporated cells were blocked in G₂/M with nocodazole (0.5 μg/mL) 32 h after electroporation for 16 h. Cells were trypsinized and resuspended in 250 μL 0.1% Triton X-100 and 50 μg/mL propidium iodide in PBS. For each sample, 10.000 GFP-expressing cells were collected by FACS (FACscan, Becton Dickinson) and analyzed for their cell cycle distribution with the CellQuest program (Becton Dickinson).

FACS analysis

Cell cycle distribution of MEFs was determined by incubating MEFs with 5-Bromodeoxyuridine (0.3 μg/mL) plus 5-Fluoro-5-deoxyuridine (0.03 μg/mL) for 5 h. Cells were collected and resuspended in 70% alcohol at 4°C, stained with DAKO mouse anti-BrdU antibody, and prepared for FACS analysis as described in Brugarolas et al. (1995).

For cell size analysis, cultures were trypsinized, fixed in 70% alcohol at 4°C, stained with propidium iodide, and analyzed by FACS as described in Brugarolas et al. (1998).

Protein analysis

Protein levels were determined by Western blot analyses following established protocols. Antibodies against pRB (C-15), p107 (C-18), p130 (C-20), p16^INK4A(M-156), CDK4 (C-22), cyclin E (M-20), p21^CIP (C-19), and p53 (FL-393) were obtained from Santa Cruz; the p19^ARF antibody (R562) was obtained from Abcam. Goat antirabbit and goat antimouse secondary antibodies were from Biosource.

Irradiation of MEFs

After 5 × 10⁵ cells were plated onto 10-cm dishes, tghey were irradiated the next day with 20 Gy of γ-radiation. Cell lysates were prepared after 24 h and analyzed on Western blots.

For cell cycle analysis, cells were irradiated with 5.5 Gy. After 14 h, cells were incubated with BrdU/FdU for 4 h and analyzed by FACS.

Soft agar assay

In GMEM containing 10% serum, 3 × 10⁴ cells were resuspended in 2 mL 0.4% low–melting point agarose (Sigma) and seeded, in duplicate, into six-well plates coated with 1% low–melting point agarose in GMEM containing 10% serum. The number of foci was scored after 2 wk.